AI Agent Hub
Back to plugins
dsh-plugin-log-forwarder preview

dsh-plugin-log-forwarder

Web Tools Updated 2026.09.09

Run the following command in DeepSeek Harness:

dsh plugin install zhaoxuejie/dsh-plugin-log-forwarder

Paste the following prompt into your AI chat to install this plugin:

Run dsh plugin install zhaoxuejie/dsh-plugin-log-forwarder in the DeepSeek Harness terminal to install the plugin; the full source repository is available at https://github.com/zhaoxuejie/dsh-plugin-log-forwarder

About this plugin

Debugging an agent is hard not because it returns a wrong answer, but because the full trace of why it took a given action evaporates the moment the session closes. dsh-plugin-log-forwarder plugs into the DeepSeek Harness session event stream, normalises every event into a single JSON shape, and fans the result out in parallel to three independent channels: a live WebSocket broadcast, a Loki push endpoint, and a local JSONL file. All three can run simultaneously, are fully decoupled from one another, and the plugin is strictly read-only, so it never interferes with the running agent and shuts down every connection cleanly on unload.

On the observability side, events can be filtered through an include-type whitelist or an exclude-type blacklist, and sensitive fields such as api_key, password, token, and secret are recursively redacted regardless of casing. Each channel surfaces live counters for forwarded, failed, and dropped events; the Loki channel backs off exponentially on network errors and marks itself disconnected after a configurable failure threshold, ready for a one-command resume. Four agent-facing tools (global status, per-channel detail, pause, resume) and a built-in side-panel UI let you inspect and control the forwarder without leaving the conversation or opening a terminal.

Three audiences will find it most useful: developers iterating on agent behaviour who need a replayable record of every reasoning step, tool call, and error; platform and SRE teams that want to funnel multiple agent sessions into Loki and surface trends or alerts in Grafana; and anyone giving a live demo who would rather point an audience at a browser watching the agent think in real time than rely on a screen recording.

Screenshots

Use Cases

  • Replay the full reasoning and tool-call trace while debugging agent behaviour
  • Funnel multi-session logs into Loki and surface trends or alerts in Grafana
  • Live-demo an agent by streaming its thinking steps to a browser in real time

Best For

  • Developers iterating on agent behaviour who need a replayable event trace
  • Platform and SRE teams centralising multi-session logs in Loki and Grafana
  • Presenters who want to stream live agent reasoning to an audience